The design language that makes your AI harness better at design.
JavaScript
Flatten the challenger draw so the same worlds stop coming back
Paul Bakaus3b35161000eb
diff --git a/skill/scripts/lib/roll-selection.mjs b/skill/scripts/lib/roll-selection.mjsindex e3c9efbb8..6fab19396 100644--- a/skill/scripts/lib/roll-selection.mjs+++ b/skill/scripts/lib/roll-selection.mjs@@ -96,31 +96,38 @@ function* rank(items, input, idFor = item => item.id) {.map(entry => entry.item);}-// Two independent exclusions, and either one is enough to hold a world back.-// Rating grades quality: a 3-star earns a second ticket, a 1-star marginal keep-// leaves the pool. Breadth says whether a world can serve an arbitrary build at-// all, so a niche world leaves however good it is, keeping its approval for-// direct briefs. Breadth was split out of rating because the only way to hold a-// narrow world back used to be calling it marginal, which made "excellent but-// narrow" unrecordable and corrupted ratings as a calibration signal.+// Rating sets how many tickets a world holds; breadth decides whether it draws+// at all. A niche world leaves the pool however good it is, keeping its approval+// for direct briefs. Breadth was split out of rating because the only way to+// hold a narrow world back used to be calling it marginal, which made "excellent+// but narrow" unrecordable and corrupted ratings as a calibration signal.+//+// Two tickets for a 3-star, one for everything else, was too sharp. Measured+// against the catalog as it stood: 3-star worlds absorbed 57% of the graphic+// draw from 65 of 163 eligible worlds, 46% of atmosphere from 13 of 43, and+// 75% of interaction from 15 of 25. The reviewer's complaint, that the same+// worlds keep coming back, is what a rating multiplier does to a pool whose+// thinnest tier holds 25 worlds.+//+// So a 3-star no longer outdraws a 2-star, and a 1-star draws at half rather+// than not at all. A marginal keep is still worth showing sometimes: the+// judgement it records is "narrow or unexceptional", not "wrong", and excluding+// it entirely made a rating do a job breadth already does properly.+const RATING_TICKETS = { 1: 1, 2: 2, 3: 2 };+const ticketsForRating = rating => RATING_TICKETS[rating] ?? 2;+function challengerTickets(pool) {return pool.flatMap(concept => {- const rating = concept.review?.rating;- if (rating === 1 || concept.review?.breadth === 'niche') return [];- return rating === 3- ? [{ concept, ticket: 0 }, { concept, ticket: 1 }]- : [{ concept, ticket: 0 }];+ if (concept.review?.breadth === 'niche') return [];+ return Array.from({ length: ticketsForRating(concept.review?.rating) },+ (_, ticket) => ({ concept, ticket }));});}function compositionTickets(pool) {- return pool.flatMap(composition => {- const rating = composition.review?.rating;- if (rating === 1) return [];- return rating === 3- ? [{ composition, ticket: 0 }, { composition, ticket: 1 }]- : [{ composition, ticket: 0 }];- });+ return pool.flatMap(composition => Array.from(+ { length: ticketsForRating(composition.review?.rating) },+ (_, ticket) => ({ composition, ticket })));}/**diff --git a/tests/concept-seed.test.mjs b/tests/concept-seed.test.mjsindex 125550ce8..2717dd849 100644--- a/tests/concept-seed.test.mjs+++ b/tests/concept-seed.test.mjs@@ -314,11 +314,19 @@ describe('concept seed scopes', () => {const graphicFirst = picks.find(pick => pick.wellTier === 'graphic');counts[graphicFirst.id] += 1;}- assert.equal(counts.marginal, 0);- // Two tickets should put the flagship on top roughly twice as often as an- // unrated peer; a generous margin keeps the assertion deterministic-safe.- assert.equal(counts.flagship > counts['solid-b'] * 1.3, true,- `flagship ${counts.flagship} vs solid-b ${counts['solid-b']}`);+ // A 1-star draws at half weight rather than not at all. Excluding it made a+ // rating do the job breadth already does, and a marginal keep records+ // "narrow or unexceptional" rather than "wrong".+ assert.equal(counts.marginal > 0, true, `marginal ${counts.marginal} should draw`);+ assert.equal(counts.marginal < counts['solid-b'], true,+ `marginal ${counts.marginal} should draw below solid-b ${counts['solid-b']}`);++ // A 3-star no longer outdraws a 2-star. The multiplier concentrated the+ // draw hard on a thin pool: measured on the live catalog, 3-star worlds took+ // 75% of the interaction draw from 15 of 25 eligible worlds.+ const spread = Math.abs(counts.flagship - counts['solid-b']) / Math.max(counts.flagship, counts['solid-b']);+ assert.equal(spread < 0.4, true,+ `flagship ${counts.flagship} and solid-b ${counts['solid-b']} should draw comparably`);// A tier holding only 1-star approvals still yields challengers.const onlyMarginal = [@@ -380,11 +388,15 @@ describe('concept seed scopes', () => {const picks = selectApprovedCompositions({ scope: 'direction', key: `stage-weight-${index}`, mode: 'persuade', sourceCompositions: pool, count: 1 });counts[picks[0].id] += 1;}- assert.equal(counts['marginal-stage'], 0, 'a 1-star composition keeps its approval but leaves the draw');- // Two tickets should put the flagship first roughly twice as often as the- // unrated peer; a generous margin keeps the assertion deterministic-safe.- assert.equal(counts['flagship-stage'] > counts['plain-stage'] * 1.3, true,- `flagship ${counts['flagship-stage']} vs plain ${counts['plain-stage']}`);+ // Same weighting as challengers: a 1-star draws at half rather than not at+ // all, and a 3-star no longer outdraws a 2-star.+ assert.equal(counts['marginal-stage'] > 0, true, 'a 1-star composition still draws, at half weight');+ assert.equal(counts['marginal-stage'] < counts['plain-stage'], true,+ `marginal ${counts['marginal-stage']} should draw below plain ${counts['plain-stage']}`);+ const stageSpread = Math.abs(counts['flagship-stage'] - counts['plain-stage'])+ / Math.max(counts['flagship-stage'], counts['plain-stage']);+ assert.equal(stageSpread < 0.4, true,+ `flagship ${counts['flagship-stage']} and plain ${counts['plain-stage']} should draw comparably`);// A pool of nothing but 1-star keeps still yields compositions.const onlyMarginal = [